Paddy from orbit: why a greener field is sometimes a worse one

1 · THE RULE THAT MATTERS MOST

An irrigated lowland paddy field peaks at heading, somewhere between 0.75 and 0.90 NDVI. Then it should drop. By ripening it belongs between 0.40 and 0.65.

That fall is the crop moving nitrogen and sugars out of the leaves and into the grain. It is the whole point of the last six weeks.

So a field still reading heading values at day 110 does not have a better canopy than its neighbours. It has delayed maturity. The usual causes are late or excess nitrogen, a variety longer than the one on record, or a late establishment date. All three move the harvest date. Two of them raise lodging risk.

Most satellite tools treat high greenness as good news at every point in the season. On paddy that is wrong for the last six weeks of it, and it is wrong in the direction that costs you money.

Of everything in our rice literature, this is the rule we would keep if we could only keep one.

2 · FOUR ECOSYSTEMS, FOUR DIFFERENT HEALTHY

Healthy NDVI for paddy is not a single curve. The same variety in an irrigated lowland field and a rainfed upland field reads differently because of water background, canopy density and how the crop was established.

We resolve bands against crop, ecosystem and phase together. Drop one and the result is not a wider band. It is no band at all.

Irrigated lowland, days after transplanting

Rainfed upland, days after sowing. Same crop, different world.

A reading of 0.55 at heading is healthy on upland and a serious problem on irrigated lowland. Rainfed lowland and flood prone deepwater have their own tables again, with the deepwater season running to 145 days.

Direct seeded fields shift every window roughly seven to ten days later than transplanted. If a field straddles two ecosystems, split it into two fields, so each one gets the band it deserves.

3 · WHAT THE SATELLITE IS ACTUALLY GOOD AT

We sort every rice problem into three groups, and we are honest about how well each one is seen from orbit.

Abiotic stress is the best detected group. Drought, salinity, submergence, poor soil. These follow topography and water, which is exactly what imagery is built to see.

Nutrient deficiency is where the platform earns its keep. NDRE responds to leaf nitrogen before symptoms are obvious to the eye.

Biotic stress needs a person. Imagery narrows down where to walk. A scout confirms what is actually there.

Satellites are excellent at abiotic and good at nutrient. Biotic detection needs ground truth, and we would rather say that than sell you a disease detector that is really a greenness detector.

4 · THE FIRST THIRTY DAYS ARE A TRAP

Early in the season, a paddy field with unusually high NDVI is usually weedy, not vigorous. The weeds are contributing the greenness while the rice stand underneath is poor.

So the alert runs backwards. Between fifteen and thirty days after transplanting we flag fields reading well above the expected baseline, not below it.

Grasses alone can cut yield by half when uncontrolled, and barnyardgrass seedlings look almost identical to young rice. The field test is the ligule: rice has one, barnyardgrass does not. Sedges give themselves away by their triangular stems.

We will also tell you plainly that weed detection is the weakest thing satellite imagery does in rice. NDVI cannot separate weed biomass from rice biomass. The first thirty days are a scouting job and imagery is confirmatory.

5 · WHEN A DROP IS NOT YET A FINDING

On flood prone fields, a sudden NDVI collapse after heavy rain is often a flood pulse rather than crop damage.

So a drop of 0.20 or more between passes opens a watch, not a finding. We raise the finding only if NDVI fails to recover on the following pass.

Water that drains leaves a canopy that climbs back. Water that stayed long enough to kill does not.

This costs us one imagery cycle of apparent responsiveness and saves you a field visit every time a bund overtops.

6 · READING THE SEASON

Different things go wrong at different points, and the yield consequences are not equal.

A few discriminators we use constantly.

NDWI drops one to two passes before NDVI does, so water stress announces itself early if you are watching the right layer. Salinity holds NDVI low even when NDWI looks healthy, because the water is there but the plant cannot use it. Nitrogen shows on older leaves first and sulfur on younger ones, and sulfur will not respond to a nitrogen top dressing, which is a diagnosis in itself.

Potassium is the subtle one. A canopy declining before the ripening window opens is a potassium finding. The same decline inside the ripening window is the band behaving normally, and we do not raise it.

7 · WHAT WE CANNOT SEE, AND SAY SO

Some of the most expensive rice problems do not show up from orbit at all. Our guides say this out loud rather than implying coverage we do not have.

False smut has no reliable satellite signature. The individual smut balls are too small. It can still take thirty to fifty percent of grains in a bad year, and the balls carry ustiloxins, which is a food safety problem as well as a yield one. Prevention is scheduled by growth stage and weather, at booting, and curative sprays afterwards do not work.

Rice bug will not appear on your dashboard. Grain damage is not canopy change. The field diagnostic is a smell. You notice it walking in.

Armyworm moves faster than the imagery cycle. Drops of 0.20 or more between passes are the signature, but a weekly satellite pass may catch it too late.

Brown planthopper hopperburn is visible, and by the time it is, damage is already severe. Circular expanding patches are one of the most distinctive biotic signatures there is, and also one of the latest.

Knowing which problems your monitoring will not catch is worth more than a dashboard that pretends to catch everything.

8 · THE LEAF COLOUR CHART PROBLEM

Two Malaysian government sources give leaf colour chart triggers for nitrogen top dressing. The Department of Agriculture's Pakej Teknologi Padi from 2008 gives 50 kg of urea per hectare in the off season and 75 in the main season. IADA Pulau Pinang's 2019 manual gives panel number 3 for direct seeded and 4 for transplanted, at 76 kg.

Neither states how many panels are on the chart. LCC charts are published in four panel and six panel versions, and number 3 is a different green on each.

A trigger panel without its chart is not a threshold. So we store the reading and which chart it came from, and we do not turn it into a verdict. It becomes evidence for the agronomist reading the record, not an input the platform acts on.

We also will not average panel numbers across a block. Panel 4 is not twice panel 2 and the steps between panels are not equal, so a mean of panel numbers is arithmetic on a rank. A block carries a distribution, not a number. SPAD readings are different and can be averaged, because a SPAD meter returns a continuous index rather than a rank.

9 · TWO SEASONS THAT ARE NOT COMPARABLE

Malaysia grows two rice crops a year, reported as main and off. In Peninsular Malaysia the main crop goes in around September and October and comes out between December and March. Sarawak runs about a month later at both ends.

The off season falls in the Southwest Monsoon and leans harder on controlled irrigation. The difference between the two seasons is structural, not performance, so a field reading lower in the off season has not necessarily got worse.

We do not offer that comparison. Compare main against main and off against off.

What we do offer, and what is genuinely useful, is a weak patch that repeats in the same place season after season. That compares a patch against the rest of the same field in the same season, which is a fair comparison and usually points at drainage, salinity or an old bund line.

10 · THE FIELD WALK

A rice field changes in days, not months, so scouting cadence is built around that. Weekly through active growth. Twice weekly from booting to heading, when blast, planthopper and stem borer all peak. Within 48 hours of a storm or a cold snap. Immediately on an alert.

Sampling is a diagonal on small fields at five stops, a double diagonal on medium at eight to ten, and a W pattern on anything large or irregular at ten to fifteen. Following an alert, a straight line through the flagged zone with at least three stops inside and two controls outside it.

At each stop, one square metre or twenty hills, and the same five steps in the same order every time. Canopy from above. Leaf colour and position. Then part the canopy and inspect the tiller base, which is where most disease and insect activity starts and where a quick walk through misses it. Panicle and grain from booting onward. Water, soil and field edge last.

One small instruction we like: walk into the wind. Insects jump away from you rather than into your face.

11 · WHAT WE WILL NOT DO

We do not check pesticide registration status. We will not refuse a scout task or a finding because of the active ingredient named in it. A named active in our guides is a description of practice, not a clearance to use it, and you should check the Department of Agriculture register before purchase. Carbofuran, long used against gall midge, has been banned in Malaysia since 2023. Butachlor is banned for rice weeds.

We do not raise an above band finding on a saturated reading. NDVI saturates above about 0.8 in every rice ecosystem, which lands inside the heading band on irrigated and rainfed lowland. Two blocks at 0.86 and 0.91 may be identical, or the second may be worse. We say the reading has saturated before we name its band.

We do not fall back to a wider band when a key is missing, and we do not average the bands that would have applied. If the ecosystem is not set or the establishment date is missing, we tell you which key is missing.